Ilker ManapandClaude Opus 5.5 4d9284ce9f Find each node's management interface, and lift the agent's memlock limit ...

The agent's defaults said eth0 before normalize looked for the interface
holding the management address, so it never looked: a node on ens3 or
bond0 had NCCL and Gloo bootstrap over an interface it does not have. It
is now found, with eth0 only as the last resort.

The agent unit sets LimitMEMLOCK=infinity. RDMA pins what it registers,
and under systemd's 8 MB default an engine without the capability that
lets root ignore the limit fails its first collective over InfiniBand.

Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com>

2026-09-29 01:44:59 +02:00

..

discovery

Fix what a multi-node launch would have got wrong on real hardware

2026-09-29 00:17:31 +02:00

engines

Fix what a multi-node launch would have got wrong on real hardware

2026-09-29 00:17:31 +02:00

provision

Move a control plane to PostgreSQL with server init --postgres

2026-09-28 14:33:03 +02:00

supervisor

Fix what a multi-node launch would have got wrong on real hardware

2026-09-29 00:17:31 +02:00

weights

Serve a model made of several files

2026-09-27 21:07:26 +02:00

agent_test.go

Find each node's management interface, and lift the agent's memlock limit

2026-09-29 01:44:59 +02:00

agent.go

Find each node's management interface, and lift the agent's memlock limit

2026-09-29 01:44:59 +02:00

failover_test.go

Let an agent find the leader instead of being told one address

2026-09-25 19:30:06 +02:00

logring_test.go

Show a node's logs on the admin page

2026-09-28 12:27:47 +02:00

logring.go

Show a node's logs on the admin page

2026-09-28 12:27:47 +02:00

mtls_test.go

Let an agent find the leader instead of being told one address

2026-09-25 19:30:06 +02:00

removeweights_test.go

Refuse a node with no room, and take back only what llmgw put there

2026-09-18 09:51:05 +02:00

selfdescribe_test.go

Install a node's engines on the node, instead of trusting what is there

2026-09-28 10:59:59 +02:00

servers_test.go

Let an agent find the leader instead of being told one address

2026-09-25 19:30:06 +02:00

shutdown_test.go

Fix an agent that could neither explain itself nor stop

2026-09-16 08:03:21 +02:00